ComfyUI Node Runs on cloud

imageOutput

Preview, save, or both β€” without breaking the image flow

By TinyTerraΒ·Created 3 years agoΒ·Updated 3 months agoΒ· 604
imageOutput
  • image
  • image
β—„image_outputPreviewβ–Ί
β—„output_path/tmp/ComfyUI/outputβ–Ί
β—„save_prefixComfyUIβ–Ί
β—„number_padding5β–Ί
β—„file_typepngβ–Ί
β—„overwrite_existingfalseβ–Ί
β—„embed_workflowtrueβ–Ί

imageOutput is the node you slap at the end of a chain when you want to see the result and keep the result, without bolting on a Save Image node that dead-ends the graph. It takes an image in, optionally saves it (with all the tinyterra filename conveniences), and passes the image through on the other side so the flow can continue - to a face detailer, an upscale pass, whatever comes next.

The name is honest: it's an output node that doesn't kill your pipeline. In the tinyterra world that matters, because their pipe samplers already have save options built in, and this is the standalone version for when you're saving an image that came from somewhere else - a RemBG result, a postprocessed frame, an intermediate you want to archive before it gets changed further downstream.

How it works

It uses the pack's shared save machinery (the same code the samplers call), so you get everything the other tinyterra save paths have:

  • image_output - the switch that matters: Hide, Preview, Save, or Hide/Save. Set it once per workflow; Preview shows without writing a file, Hide/Save writes quietly without spamming the preview pane.
  • output_path and save_prefix - where and what to name it. The prefix supports the pack's %date:yyyy-MM-dd-hh-mm-ss% syntax and %widget%-style value insertion, so you can bake the date or an upstream setting into filenames.
  • number_padding - how many digits in the counter suffix (default 5, so _00001).
  • file_type - png, jpg, jpeg, tiff, tif, webp, or bmp.
  • overwrite_existing - off by default, so repeated runs roll the counter instead of clobbering.
  • embed_workflow - on by default, which bakes the workflow JSON into the PNG so you can drag the image back into ComfyUI later. This is the feature you want on for anything you might share.

The image output on the right passes the original pixels through untouched - that's the throughput that makes this node a "writer with a bypass" rather than a dead end.

Installing it

It ships with the whole tinyterraNodes pack:

cd ComfyUI/custom_nodes
git clone https://github.com/TinyTerra/ComfyUI_tinyterraNodes.git

or ComfyUI Manager β†’ tinyterraNodes. No dependencies, no model files.

Gotchas

Two things to keep in mind. First, image_output isn't per-image - it's a graph-level choice. If you set Save you're committing to files on every run, so pairs like Preview while iterating and Hide/Save once you're happy are the sensible pattern. Second, if you're chaining this into another image node, remember the passed-through image is the input image, not a re-encoded copy - which is exactly what you want, but worth knowing so you don't expect it to normalize formats. And if all you need is a quick look at a sampler's output, the pipe samplers' built-in image_output widget does the same thing one node earlier.

Category🌏 tinyterra/image

Inputs (8)

NameTypeDefaultDescription
imageIMAGEβ€”
image_outputCOMBOPreview4 options: Hide, Preview, Save, Hide/Save
output_pathSTRING/tmp/ComfyUI/outputβ€”
save_prefixSTRINGComfyUIβ€”
number_paddingCOMBO59 options: None, 2, 3, 4, 5, 6, +3
file_typeCOMBOpng7 options: png, jpg, jpeg, tiff, tif, webp, +1
overwrite_existingBOOLEANfalseβ€”
embed_workflowBOOLEANtrueβ€”

Outputs (1)

NameTypeDescription
imageIMAGEβ€”